Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Found mnist Data
Found celeba Data

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/matplotlib/__init__.py:1067: UserWarning: Duplicate key in file "/home/ec2-user/.config/matplotlib/matplotlibrc", line #2
  (fname, cnt))
/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/matplotlib/__init__.py:1067: UserWarning: Duplicate key in file "/home/ec2-user/.config/matplotlib/matplotlibrc", line #3
  (fname, cnt))

<matplotlib.image.AxesImage at 0x7f3eacf438d0>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7f3eace62f28>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

/home/ec2-user/anaconda3/envs/tensorflow_p36/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters

TensorFlow Version: 1.6.0
Default GPU Device: /device:GPU:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    input_real = tf.placeholder(tf.float32, [None, image_width, image_height, image_channels], "input_real")
    input_z = tf.placeholder(tf.float32, [None, z_dim], "input_z")
    learning_rate = tf.placeholder(tf.float32, None, "learning_rate")

    return input_real, input_z, learning_rate


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

def discriminator(images, reuse=False, alpha=0.01):
    """
    Create the discriminator network
    :param image: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    with tf.variable_scope('discriminator', reuse=reuse):
        x1 = tf.layers.conv2d(images, 64, 5, strides=2, padding="same", kernel_initializer=tf.contrib.layers.xavier_initializer())
        relu1 = tf.maximum(alpha * x1, x1)

        x2 = tf.layers.conv2d(relu1, 128, 5, strides=2, padding="same", kernel_initializer=tf.contrib.layers.xavier_initializer())
        bn2 = tf.layers.batch_normalization(x2, training=True)
        relu2 = tf.maximum(alpha * bn2, bn2)
        d2 = tf.nn.dropout(relu2, 0.8)

        x3 = tf.layers.conv2d(d2, 256, 5, strides=2, padding="same", kernel_initializer=tf.contrib.layers.xavier_initializer())
        bn3 = tf.layers.batch_normalization(x3, training=True)
        relu3 = tf.maximum(alpha * bn3, bn3)
        d3 = tf.nn.dropout(relu3, 0.8)

        flat = tf.reshape(d3, (-1, 4*4*256))
        logits = tf.layers.dense(flat, 1)
        out = tf.sigmoid(logits)

    return out, logits

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z, out_channel_dim, is_train=True, alpha=0.01):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
    with tf.variable_scope("generator", reuse=not is_train):
        x1 = tf.layers.dense(z, 4*4*256)
        x1 = tf.reshape(x1, (-1, 4, 4, 256))
        x1 = tf.layers.dense(z, 7*7*512)
        x1 = tf.reshape(x1, (-1, 7, 7, 512))
        x1 = tf.layers.batch_normalization(x1, training=is_train)
        relu1 = tf.maximum(alpha * x1, x1)
        
        x2 = tf.layers.conv2d_transpose(relu1, 256, 5, 2, 'same', kernel_initializer=tf.contrib.layers.xavier_initializer())
        bn2 = tf.layers.batch_normalization(x2, training=is_train)
        relu2 = tf.maximum(alpha * bn2, bn2)
        d2 = tf.nn.dropout(relu2, 0.8)
        
        x3 = tf.layers.conv2d_transpose(x2, 128, 5, 2, 'same', kernel_initializer=tf.contrib.layers.xavier_initializer())
        bn3 = tf.layers.batch_normalization(x3, training=is_train)
        relu3 = tf.maximum(alpha * bn3, bn3)
        d3 = tf.nn.dropout(relu3, 0.8)
        
        logits = tf.layers.conv2d_transpose(d3, out_channel_dim, 5, 1, 'same')
        out = tf.tanh(logits)

        return out


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim, alpha=0.9):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    g_model = generator(input_z, out_channel_dim)
    d_model_real, d_logits_real = discriminator(input_real)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse=True)

    d_loss_real = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_real, labels=tf.ones_like(d_logits_real) * alpha))
    d_loss_fake = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.zeros_like(d_logits_fake)))
    g_loss = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.ones_like(d_logits_fake)))

    d_loss = d_loss_real + d_loss_fake

    return d_loss, g_loss


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)

    return d_train_opt, g_train_opt


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    input_real, input_z, lr = model_inputs(data_shape[1], data_shape[2], data_shape[3], z_dim)
    d_loss, g_loss = model_loss(input_real, input_z, data_shape[3])
    d_opt, g_opt = model_opt(d_loss, g_loss, lr, beta1)
    dis, gen, f1 = [], [], []
    
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        for epoch_i in range(epoch_count):
            steps = 0
            for batch_images in get_batches(batch_size):
                steps +=1
                batch_images = batch_images * 2
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))
                _ = sess.run(d_opt, feed_dict={input_real: batch_images, input_z: batch_z, lr: learning_rate})
                _ = sess.run(g_opt, feed_dict={input_real: batch_images, input_z: batch_z, lr: learning_rate})
                train_loss_d = d_loss.eval({input_real: batch_images, input_z: batch_z})
                train_loss_g = g_loss.eval({input_z: batch_z})
                dis.append(train_loss_d)
                gen.append(train_loss_g)
                val = (train_loss_d + train_loss_g) / 2
                f1.append(val)
                if steps % 20 == 0:

                    print("Epoch {}/{}...".format(epoch_i+1, epochs),
                          "Batch {}...".format(steps),
                          "Discriminator Loss: {:.4f}...".format(train_loss_d),
                          "Generator Loss: {:.4f}".format(train_loss_g))

                if steps % 200 == 0:
                    show_generator_output(sess, show_n_images, input_z, data_shape[3], data_image_mode)
    return dis, gen, f1

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 64
z_dim = 164
learning_rate = 0.0002
beta1 = 0.4


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch 1/2... Batch 20... Discriminator Loss: 0.6456... Generator Loss: 1.6505
Epoch 1/2... Batch 40... Discriminator Loss: 1.0095... Generator Loss: 1.1159
Epoch 1/2... Batch 60... Discriminator Loss: 1.3518... Generator Loss: 0.7359
Epoch 1/2... Batch 80... Discriminator Loss: 2.0054... Generator Loss: 0.2956
Epoch 1/2... Batch 100... Discriminator Loss: 1.5965... Generator Loss: 0.7283
Epoch 1/2... Batch 120... Discriminator Loss: 1.3091... Generator Loss: 0.9094
Epoch 1/2... Batch 140... Discriminator Loss: 1.5384... Generator Loss: 0.5098
Epoch 1/2... Batch 160... Discriminator Loss: 1.3613... Generator Loss: 1.6343
Epoch 1/2... Batch 180... Discriminator Loss: 0.9609... Generator Loss: 1.0689
Epoch 1/2... Batch 200... Discriminator Loss: 1.0003... Generator Loss: 1.1790

Epoch 1/2... Batch 220... Discriminator Loss: 1.1106... Generator Loss: 1.5356
Epoch 1/2... Batch 240... Discriminator Loss: 1.0232... Generator Loss: 1.1826
Epoch 1/2... Batch 260... Discriminator Loss: 1.2529... Generator Loss: 0.8279
Epoch 1/2... Batch 280... Discriminator Loss: 1.1058... Generator Loss: 0.9399
Epoch 1/2... Batch 300... Discriminator Loss: 1.4958... Generator Loss: 0.5481
Epoch 1/2... Batch 320... Discriminator Loss: 1.2367... Generator Loss: 0.7501
Epoch 1/2... Batch 340... Discriminator Loss: 1.1971... Generator Loss: 0.9877
Epoch 1/2... Batch 360... Discriminator Loss: 1.2197... Generator Loss: 0.9340
Epoch 1/2... Batch 380... Discriminator Loss: 1.1784... Generator Loss: 1.4323
Epoch 1/2... Batch 400... Discriminator Loss: 1.3588... Generator Loss: 1.7363

Epoch 1/2... Batch 420... Discriminator Loss: 1.5695... Generator Loss: 0.4599
Epoch 1/2... Batch 440... Discriminator Loss: 1.1679... Generator Loss: 1.6431
Epoch 1/2... Batch 460... Discriminator Loss: 1.1289... Generator Loss: 1.0293
Epoch 1/2... Batch 480... Discriminator Loss: 1.3238... Generator Loss: 0.7796
Epoch 1/2... Batch 500... Discriminator Loss: 1.0608... Generator Loss: 0.9604
Epoch 1/2... Batch 520... Discriminator Loss: 1.2731... Generator Loss: 0.7555
Epoch 1/2... Batch 540... Discriminator Loss: 1.1920... Generator Loss: 1.1739
Epoch 1/2... Batch 560... Discriminator Loss: 0.9281... Generator Loss: 1.0921
Epoch 1/2... Batch 580... Discriminator Loss: 1.1812... Generator Loss: 0.9493
Epoch 1/2... Batch 600... Discriminator Loss: 1.1798... Generator Loss: 1.6398

Epoch 1/2... Batch 620... Discriminator Loss: 1.2609... Generator Loss: 0.7447
Epoch 1/2... Batch 640... Discriminator Loss: 1.1444... Generator Loss: 1.0963
Epoch 1/2... Batch 660... Discriminator Loss: 1.2582... Generator Loss: 1.8242
Epoch 1/2... Batch 680... Discriminator Loss: 1.2858... Generator Loss: 0.6725
Epoch 1/2... Batch 700... Discriminator Loss: 1.1767... Generator Loss: 1.3288
Epoch 1/2... Batch 720... Discriminator Loss: 0.9782... Generator Loss: 1.3712
Epoch 1/2... Batch 740... Discriminator Loss: 1.2448... Generator Loss: 0.6277
Epoch 1/2... Batch 760... Discriminator Loss: 1.0728... Generator Loss: 1.8055
Epoch 1/2... Batch 780... Discriminator Loss: 1.6512... Generator Loss: 0.4734
Epoch 1/2... Batch 800... Discriminator Loss: 1.2611... Generator Loss: 0.6661

Epoch 1/2... Batch 820... Discriminator Loss: 1.0527... Generator Loss: 1.3956
Epoch 1/2... Batch 840... Discriminator Loss: 1.0931... Generator Loss: 1.0458
Epoch 1/2... Batch 860... Discriminator Loss: 1.3596... Generator Loss: 0.6532
Epoch 1/2... Batch 880... Discriminator Loss: 1.2803... Generator Loss: 0.6406
Epoch 1/2... Batch 900... Discriminator Loss: 1.2512... Generator Loss: 0.7788
Epoch 1/2... Batch 920... Discriminator Loss: 1.1501... Generator Loss: 0.8651
Epoch 2/2... Batch 20... Discriminator Loss: 0.9164... Generator Loss: 1.2500
Epoch 2/2... Batch 40... Discriminator Loss: 0.9070... Generator Loss: 1.3530
Epoch 2/2... Batch 60... Discriminator Loss: 1.0797... Generator Loss: 0.9796
Epoch 2/2... Batch 80... Discriminator Loss: 1.2874... Generator Loss: 0.7395
Epoch 2/2... Batch 100... Discriminator Loss: 1.0812... Generator Loss: 1.5130
Epoch 2/2... Batch 120... Discriminator Loss: 1.1413... Generator Loss: 0.7963
Epoch 2/2... Batch 140... Discriminator Loss: 1.2440... Generator Loss: 0.7391
Epoch 2/2... Batch 160... Discriminator Loss: 1.1417... Generator Loss: 0.9166
Epoch 2/2... Batch 180... Discriminator Loss: 1.3496... Generator Loss: 0.6532
Epoch 2/2... Batch 200... Discriminator Loss: 0.9149... Generator Loss: 1.3633

Epoch 2/2... Batch 220... Discriminator Loss: 0.8845... Generator Loss: 1.9784
Epoch 2/2... Batch 240... Discriminator Loss: 1.0755... Generator Loss: 1.0268
Epoch 2/2... Batch 260... Discriminator Loss: 0.9950... Generator Loss: 1.0075
Epoch 2/2... Batch 280... Discriminator Loss: 1.1054... Generator Loss: 0.8298
Epoch 2/2... Batch 300... Discriminator Loss: 1.2368... Generator Loss: 0.7643
Epoch 2/2... Batch 320... Discriminator Loss: 0.9886... Generator Loss: 1.0735
Epoch 2/2... Batch 340... Discriminator Loss: 0.9186... Generator Loss: 1.1207
Epoch 2/2... Batch 360... Discriminator Loss: 1.0056... Generator Loss: 1.0889
Epoch 2/2... Batch 380... Discriminator Loss: 1.2531... Generator Loss: 0.7552
Epoch 2/2... Batch 400... Discriminator Loss: 0.9462... Generator Loss: 2.0088

Epoch 2/2... Batch 420... Discriminator Loss: 0.9422... Generator Loss: 1.5262
Epoch 2/2... Batch 440... Discriminator Loss: 0.8873... Generator Loss: 1.9585
Epoch 2/2... Batch 460... Discriminator Loss: 0.8215... Generator Loss: 1.5885
Epoch 2/2... Batch 480... Discriminator Loss: 0.9618... Generator Loss: 1.6660
Epoch 2/2... Batch 500... Discriminator Loss: 1.1627... Generator Loss: 0.9030
Epoch 2/2... Batch 520... Discriminator Loss: 1.1555... Generator Loss: 0.8804
Epoch 2/2... Batch 540... Discriminator Loss: 0.9578... Generator Loss: 1.3067
Epoch 2/2... Batch 560... Discriminator Loss: 0.9110... Generator Loss: 1.4129
Epoch 2/2... Batch 580... Discriminator Loss: 1.2851... Generator Loss: 0.9313
Epoch 2/2... Batch 600... Discriminator Loss: 0.9986... Generator Loss: 1.1369

Epoch 2/2... Batch 620... Discriminator Loss: 1.0612... Generator Loss: 0.9311
Epoch 2/2... Batch 640... Discriminator Loss: 0.9896... Generator Loss: 1.8075
Epoch 2/2... Batch 660... Discriminator Loss: 0.8693... Generator Loss: 1.2803
Epoch 2/2... Batch 680... Discriminator Loss: 1.0398... Generator Loss: 0.9434
Epoch 2/2... Batch 700... Discriminator Loss: 0.8987... Generator Loss: 1.3717
Epoch 2/2... Batch 720... Discriminator Loss: 0.8563... Generator Loss: 1.4728
Epoch 2/2... Batch 740... Discriminator Loss: 1.0961... Generator Loss: 1.0486
Epoch 2/2... Batch 760... Discriminator Loss: 1.0187... Generator Loss: 1.0737
Epoch 2/2... Batch 780... Discriminator Loss: 1.2312... Generator Loss: 0.7989
Epoch 2/2... Batch 800... Discriminator Loss: 0.8062... Generator Loss: 1.6167

Epoch 2/2... Batch 820... Discriminator Loss: 1.1153... Generator Loss: 2.3253
Epoch 2/2... Batch 840... Discriminator Loss: 0.8812... Generator Loss: 1.3766
Epoch 2/2... Batch 860... Discriminator Loss: 1.0888... Generator Loss: 0.9276
Epoch 2/2... Batch 880... Discriminator Loss: 0.9426... Generator Loss: 1.2359
Epoch 2/2... Batch 900... Discriminator Loss: 0.8302... Generator Loss: 1.8483
Epoch 2/2... Batch 920... Discriminator Loss: 0.7848... Generator Loss: 1.6761

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 64
z_dim = 164
learning_rate = 0.0002
beta1 = 0.4


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 10

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    dis, gen, f1 = train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch 1/10... Batch 20... Discriminator Loss: 0.5552... Generator Loss: 2.4103
Epoch 1/10... Batch 40... Discriminator Loss: 0.3912... Generator Loss: 4.2242
Epoch 1/10... Batch 60... Discriminator Loss: 1.8449... Generator Loss: 0.5678
Epoch 1/10... Batch 80... Discriminator Loss: 0.5326... Generator Loss: 3.2501
Epoch 1/10... Batch 100... Discriminator Loss: 0.4207... Generator Loss: 4.3603
Epoch 1/10... Batch 120... Discriminator Loss: 0.5147... Generator Loss: 2.6756
Epoch 1/10... Batch 140... Discriminator Loss: 0.4301... Generator Loss: 4.6150
Epoch 1/10... Batch 160... Discriminator Loss: 2.8393... Generator Loss: 10.0612
Epoch 1/10... Batch 180... Discriminator Loss: 0.4993... Generator Loss: 3.7260
Epoch 1/10... Batch 200... Discriminator Loss: 0.4641... Generator Loss: 3.1522

Epoch 1/10... Batch 220... Discriminator Loss: 0.4865... Generator Loss: 3.0062
Epoch 1/10... Batch 240... Discriminator Loss: 0.4694... Generator Loss: 3.2220
Epoch 1/10... Batch 260... Discriminator Loss: 0.5715... Generator Loss: 2.2278
Epoch 1/10... Batch 280... Discriminator Loss: 0.4819... Generator Loss: 2.6776
Epoch 1/10... Batch 300... Discriminator Loss: 0.6986... Generator Loss: 2.0415
Epoch 1/10... Batch 320... Discriminator Loss: 0.7881... Generator Loss: 2.1672
Epoch 1/10... Batch 340... Discriminator Loss: 0.6996... Generator Loss: 1.6772
Epoch 1/10... Batch 360... Discriminator Loss: 1.5986... Generator Loss: 0.5239
Epoch 1/10... Batch 380... Discriminator Loss: 0.7499... Generator Loss: 1.4270
Epoch 1/10... Batch 400... Discriminator Loss: 0.8455... Generator Loss: 1.3728

Epoch 1/10... Batch 420... Discriminator Loss: 1.1836... Generator Loss: 0.7597
Epoch 1/10... Batch 440... Discriminator Loss: 0.6608... Generator Loss: 2.2006
Epoch 1/10... Batch 460... Discriminator Loss: 1.1255... Generator Loss: 0.9460
Epoch 1/10... Batch 480... Discriminator Loss: 0.8275... Generator Loss: 4.1413
Epoch 1/10... Batch 500... Discriminator Loss: 1.0076... Generator Loss: 1.0130
Epoch 1/10... Batch 520... Discriminator Loss: 0.8414... Generator Loss: 1.3625
Epoch 1/10... Batch 540... Discriminator Loss: 1.1146... Generator Loss: 3.1425
Epoch 1/10... Batch 560... Discriminator Loss: 0.8430... Generator Loss: 1.2221
Epoch 1/10... Batch 580... Discriminator Loss: 1.0226... Generator Loss: 2.8060
Epoch 1/10... Batch 600... Discriminator Loss: 0.6606... Generator Loss: 2.2839

Epoch 1/10... Batch 620... Discriminator Loss: 1.1024... Generator Loss: 0.8055
Epoch 1/10... Batch 640... Discriminator Loss: 0.7400... Generator Loss: 1.8338
Epoch 1/10... Batch 660... Discriminator Loss: 1.3697... Generator Loss: 0.5812
Epoch 1/10... Batch 680... Discriminator Loss: 1.0634... Generator Loss: 0.8923
Epoch 1/10... Batch 700... Discriminator Loss: 1.1428... Generator Loss: 0.7548
Epoch 1/10... Batch 720... Discriminator Loss: 0.9107... Generator Loss: 2.2520
Epoch 1/10... Batch 740... Discriminator Loss: 1.0394... Generator Loss: 2.2906
Epoch 1/10... Batch 760... Discriminator Loss: 0.7949... Generator Loss: 1.5171
Epoch 1/10... Batch 780... Discriminator Loss: 0.9193... Generator Loss: 1.0526
Epoch 1/10... Batch 800... Discriminator Loss: 0.8045... Generator Loss: 2.5132

Epoch 1/10... Batch 820... Discriminator Loss: 0.9157... Generator Loss: 2.0890
Epoch 1/10... Batch 840... Discriminator Loss: 0.7494... Generator Loss: 1.5601
Epoch 1/10... Batch 860... Discriminator Loss: 0.7715... Generator Loss: 2.1017
Epoch 1/10... Batch 880... Discriminator Loss: 1.5350... Generator Loss: 0.4148
Epoch 1/10... Batch 900... Discriminator Loss: 1.0732... Generator Loss: 1.1044
Epoch 1/10... Batch 920... Discriminator Loss: 0.9417... Generator Loss: 1.9158
Epoch 1/10... Batch 940... Discriminator Loss: 1.0016... Generator Loss: 1.3013
Epoch 1/10... Batch 960... Discriminator Loss: 1.0775... Generator Loss: 0.9948
Epoch 1/10... Batch 980... Discriminator Loss: 1.7278... Generator Loss: 0.4033
Epoch 1/10... Batch 1000... Discriminator Loss: 1.0529... Generator Loss: 1.3000

Epoch 1/10... Batch 1020... Discriminator Loss: 1.4964... Generator Loss: 0.5490
Epoch 1/10... Batch 1040... Discriminator Loss: 1.1991... Generator Loss: 0.9291
Epoch 1/10... Batch 1060... Discriminator Loss: 1.1864... Generator Loss: 1.1830
Epoch 1/10... Batch 1080... Discriminator Loss: 1.4607... Generator Loss: 0.6161
Epoch 1/10... Batch 1100... Discriminator Loss: 1.2913... Generator Loss: 1.0110
Epoch 1/10... Batch 1120... Discriminator Loss: 1.2767... Generator Loss: 0.7034
Epoch 1/10... Batch 1140... Discriminator Loss: 1.1922... Generator Loss: 1.1539
Epoch 1/10... Batch 1160... Discriminator Loss: 1.2105... Generator Loss: 1.0051
Epoch 1/10... Batch 1180... Discriminator Loss: 1.4490... Generator Loss: 0.7520
Epoch 1/10... Batch 1200... Discriminator Loss: 1.6365... Generator Loss: 0.5173

Epoch 1/10... Batch 1220... Discriminator Loss: 1.4132... Generator Loss: 0.6518
Epoch 1/10... Batch 1240... Discriminator Loss: 1.7169... Generator Loss: 0.3853
Epoch 1/10... Batch 1260... Discriminator Loss: 1.4547... Generator Loss: 0.6868
Epoch 1/10... Batch 1280... Discriminator Loss: 1.1809... Generator Loss: 0.9151
Epoch 1/10... Batch 1300... Discriminator Loss: 1.4424... Generator Loss: 0.6425
Epoch 1/10... Batch 1320... Discriminator Loss: 1.3581... Generator Loss: 0.9942
Epoch 1/10... Batch 1340... Discriminator Loss: 1.3795... Generator Loss: 0.9668
Epoch 1/10... Batch 1360... Discriminator Loss: 1.4906... Generator Loss: 0.6345
Epoch 1/10... Batch 1380... Discriminator Loss: 1.2300... Generator Loss: 0.9891
Epoch 1/10... Batch 1400... Discriminator Loss: 1.4912... Generator Loss: 0.4438

Epoch 1/10... Batch 1420... Discriminator Loss: 1.4073... Generator Loss: 1.0224
Epoch 1/10... Batch 1440... Discriminator Loss: 1.4976... Generator Loss: 0.6830
Epoch 1/10... Batch 1460... Discriminator Loss: 1.3815... Generator Loss: 0.6915
Epoch 1/10... Batch 1480... Discriminator Loss: 1.2719... Generator Loss: 1.2238
Epoch 1/10... Batch 1500... Discriminator Loss: 1.4473... Generator Loss: 0.5926
Epoch 1/10... Batch 1520... Discriminator Loss: 1.1424... Generator Loss: 0.9013
Epoch 1/10... Batch 1540... Discriminator Loss: 1.0724... Generator Loss: 1.0371
Epoch 1/10... Batch 1560... Discriminator Loss: 1.2403... Generator Loss: 0.7472
Epoch 1/10... Batch 1580... Discriminator Loss: 1.2084... Generator Loss: 0.9321
Epoch 1/10... Batch 1600... Discriminator Loss: 1.3334... Generator Loss: 0.6880

Epoch 1/10... Batch 1620... Discriminator Loss: 1.4445... Generator Loss: 0.5921
Epoch 1/10... Batch 1640... Discriminator Loss: 1.4616... Generator Loss: 0.6613
Epoch 1/10... Batch 1660... Discriminator Loss: 1.3079... Generator Loss: 0.6229
Epoch 1/10... Batch 1680... Discriminator Loss: 1.2227... Generator Loss: 0.8343
Epoch 1/10... Batch 1700... Discriminator Loss: 1.3576... Generator Loss: 0.8829
Epoch 1/10... Batch 1720... Discriminator Loss: 1.3737... Generator Loss: 0.7809
Epoch 1/10... Batch 1740... Discriminator Loss: 1.4931... Generator Loss: 0.8993
Epoch 1/10... Batch 1760... Discriminator Loss: 1.1466... Generator Loss: 1.1428
Epoch 1/10... Batch 1780... Discriminator Loss: 1.2694... Generator Loss: 0.9236
Epoch 1/10... Batch 1800... Discriminator Loss: 1.3533... Generator Loss: 0.6052

Epoch 1/10... Batch 1820... Discriminator Loss: 1.3242... Generator Loss: 0.8150
Epoch 1/10... Batch 1840... Discriminator Loss: 1.4723... Generator Loss: 0.4992
Epoch 1/10... Batch 1860... Discriminator Loss: 1.2364... Generator Loss: 0.8090
Epoch 1/10... Batch 1880... Discriminator Loss: 1.3279... Generator Loss: 0.6943
Epoch 1/10... Batch 1900... Discriminator Loss: 1.2764... Generator Loss: 0.8493
Epoch 1/10... Batch 1920... Discriminator Loss: 1.2762... Generator Loss: 0.8601
Epoch 1/10... Batch 1940... Discriminator Loss: 1.4757... Generator Loss: 0.7193
Epoch 1/10... Batch 1960... Discriminator Loss: 1.6623... Generator Loss: 0.4839
Epoch 1/10... Batch 1980... Discriminator Loss: 1.4053... Generator Loss: 0.7206
Epoch 1/10... Batch 2000... Discriminator Loss: 1.6765... Generator Loss: 0.5070

import matplotlib.pyplot as plt

plt.plot(dis)
plt.show()
plt.plot(gen)
plt.show()
plt.plot(f1)
plt.show()

idx = f1.index(min(f1))
print(dis[idx], gen[idx])

0.83051276 0.5755906

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.